The Impact of the Fourth Industrial Revolution on Jobs: An Analysis of the Swiss Financial Services Industry

Abstract

Inequality, Polarization

Financial Services many technologies affecting and changing employment, staff cuts of 10%

Most important transformation of jobs, even if number constant

Biggest factor education, retraining during lifetime

Yes, automation and robotics will bring advances and benefits to people — but only a select few. Shareholders, top earners, and the well-educated will enjoy most of the benefits that come from increased corporate productivity and a demand for technical, highly-skilled roles.

Table of Contents

Abstract

Abbreviations

Literaturverzeichnis

1 Introduction

2 The Fourth Industrial Revolution

2.1 A Brief History

2.2 Why Could This Time Be Different?

2.3 Some Context: Technological Drivers

3 Impact of the Fourth Industrial Revolution on jobs

3.1 Perspective A: Unemployment will increase dramatically

3.2 Perspective B: Substituted jobs will be replaced with new ones

3.3 Representation of opinions in the media

3.4 Consolidation of opinions

3.5 Skills demand will change

4 How to leverage automation as a society

4.1 Measures related to education

4.2 New incentives

4.3 Welfare-related measures

5. Excursus: Why would a jobless future be bad?

6. The Swiss financial services industry

6.1 Industry Overview

3.2 Disruptions and innovations in financial services

1 Introduction

Fears of automation taking over our jobs and making us redundant have a long history, going back to the First Industrial Revolution, when textile workers – the Luddites – rioted out of concern for their livelihoods. Since then, the debate about technological unemployment, a term coined by John Maynard Keynes in the 1930s, reemerged every few decades (Mokyr, Vickers, & Ziebarth, 2015, p. 34). In an essay, in which he imagined the world in 2030, Keynes predicted that “due to our discovery of means of economizing the use of labor outrunning the pace at which we can find new uses for labor” we would eventually reach the 15-hour work week (Keynes, 1930, p. 3).

Thus, today’s concerns fueled by scientific breakthroughs in fields such as machine learning, robotics, and artificial intelligence, are not new but have been around for 200 years. However, as historian and philosopher Yuval Noah Harari put it, this might be a matter of the boy who cried wolf eventually being right (Harari, 2017, Position 665). Just because these fears have not proven to be true in the past, with new jobs outnumbering the ones impacted by emerging technologies, the accelerating pace of technological development in automation and digitalization might bring about a jobless future (Amtz, Gregory, & Zierahn, 2016, p. 7). A widely-cited 2013 report by Osborne and Frey suggested that 47% of jobs in the United States could be automatable in the next few decades (Frey & Osborne, 2013, p. 38). With impressive breakthroughs, such as Watson winning Jeopardy, or self-driving cars being rolled out – things that were regarded skeptically only a decade ago – it is a good bet that we will be surprised by progress in the coming decades.

In bestselling book “The Second Machine Age”, the MIT scientists Brynjolfsson and McAfee argue that due to the exponential development of technology we will have to adapt and learn how to work or even race with the machines sooner rather than later. Martin Ford goes even further in “Rise of the Robots” and warns that as a society we must radically rethink the way we distribute wealth and makes a case for the introduction of a Basic Income. He suggests that a massive paradigm shift is under way in how we view technology. Until now machines were tools increasing our productivity – in the future they might themselves turn into workers and replace us (Ford, 2015, p. xii). Publications like “Rise of the Robots” and “The Second Machine Age” have incited and stoked a public dialogue on the topic of rising inequality, job polarization, and ways to manage the changing skill requirements of our workforce. We have already been experiencing rising inequality, falling real wages and a declining share of income flowing to workers as opposed to capital, and these trends are likely to continue (Avent, 2016, p. 3).

The World Economic Forum argues that we are at the beginning of the Fourth Industrial Revolution, which will blur the lines between the physical, biological, and physical world at an unprecedented scale, speed and impact. This would profoundly change our society, causing shifts between industries, enabling new business models and reshaping the way we communicate and work (Schwab, 2016, p. v).

However, there is a multitude of experts who disagree on the impact that these changes will have on employment. A widely-cited example on the topic is the shift of employment from agriculture. In 1800, 66% of Switzerland’s population worked in agriculture, compared with only 3% today (Deloitte, 2016, p. 3). This of course did not lead to a reduction in employment but just meant that jobs were created in other industries. According to a Deloitte report, in the last 25 years, 800’000 new jobs were created in Switzerland despite automation and recent projections predict that until 2025 around 270’000 new jobs will be created (Deloitte, 2016, p. 1). The authors argue that the creation of new jobs will outweigh the substitution effect caused by automation, as it always has in the past. Automating a certain task increases productivity, which should in turn increase the demand for human workers doing other tasks around it that haven’t been automated yet. This is called the complementary effect (Autor, 2015, p. 5).

As the disagreement between experts shows, the question is whether the substitution or complementary effect will prevail. This thesis will take up this question with specific regard to the financial services industry in Switzerland. Therefore, the first part of this paper will analyze several current viewpoints on the effects of automation on labor globally. It will also look at approaches of how society could respond to these challenges. In the second part, it will focus on the financial services industry in Switzerland and analyze the impact of disruptions on jobs and needed skills. Specifically, it will aim to answer the question if the number of jobs in the financial services industry will radically decline because of new technologies, or stay the same respectively even increase. But first, it is necessary to define the characteristics of the Fourth Industrial Revolution and see if it is truly different from what preceded it.

2 The Fourth Industrial Revolution

2.1 A Brief History

Before delving into the specifics of the Fourth Industrial Revolution, let us go into what brought us to this point. A revolution is defined as abrupt and radical change. The first radical change in how humankind lived happened 10’000 years ago with the settlement of humans and domestication of animals. This was called the Agricultural Revolution. After some incremental changes in the following ten millennia, the next radical shift was the First Industrial Revolution, or what is typically understood when talking about the Industrial Revolution. This started in the second half of the 18^th century and kicked off almost three centuries of accelerating societal and technological shifts. It was characterized by the change from muscle power to the steam engine and gave rise to the steel and textile industry as well as the train as means of transportation. The transition period lasted until the late 19^th century, when the Second Industrial Revolution set in with the development of methods for mass production with electricity as a driving force (Prisecaru, 2016, p. 57). Industrial firms combined costly capital equipment with human labor and reorganized the process of production to maximize its output. This necessitated an assembly line production with every worker performing a few repetitive tasks, following tight schedules and strict discipline. Jobs were wholly transformed and humans became cogs in a machine. However, with machines becoming more sophisticated, the demand for highly skilled workers grew. Thus, in the 20^th century the number of people getting a post-secondary education grew steadily. Workers found themselves in a “race between education and technology” as Claudia Goldin, a Harvard University economist, put it (The Economist, 2014). This of course sounds very familiar to what is often said about the situation today.

The Third Industrial Revolution began in the 1960s and is also referred to as the Computer or Digital Revolution. It was prompted by the development of personal computing, mainframe computing, the development of semiconductors, and the internet (Schwab, 2016, p. 5). This once again transformed the nature of work, making it possible to automate highly routine jobs such as the production-line jobs created in the Second Industrial Revolution. The share of employment in the manufacturing sector in the US declined from almost 30% in the 1950s to less than 10% while the percentage of people working in the services industry increased to almost 70%. Thus, skilled labor became more important than ever (The Economist, 2014).

In the last decade, we have now entered the Fourth Industrial Revolution, which is leading to the augmentation of human production with enhanced cognitive power. This Fourth Industrial Revolution is based on emerging technologies such as Artificial Intelligence, machine learning, 3D printing, the Internet of Things, and genetic engineering (Schwab, 2016, p. 5). In Germany, Industry 4.0 was coined at the 2011 Hannover fair and describes how automation will revolutionize global value chains and create smart factories (Schwab, 2016, p. 6) The term Fourth Industrial Revolution has especially gained prominence in the last couple of years, when the World Economic Forum did much to popularize it, making “Mastering the Fourth Industrial Revolution” the main topic of the Annual Meeting in Davos 2016. Looking at the Google Trends search history, one can see the interest spiking around Davos 2016 after a decade of little to no interest, and then increasing again leading up to Davos 2017 (Google Trends). In the past, these technological changes were seen as part of the Third Industrial Revolution, or referred to with other terms such as the Second Machine Age. But now the term Fourth Industrial Revolution seems to have established itself. But whether this phase will be regarded as an extension of the Third Industrial Revolution in the future or as a new Revolution, as the World Economic Forum argues, it cannot be denied that it will have a lasting impact on society.

To summarize, we have seen an evolution from a farming society to today’s capitalist system, gradually moving the labor force from physical to mental activity. The question is whether the technological progress and productivity growth accompanying this development will lead to increased wealth, like it has in the past, or increase unemployment and social inequality. To examine this issue, let us first look at whether this time will be different from the societal shifts in the past.

2.2 Why Could This Time Be Different?

As mentioned in the introduction, experts disagree on the extent of the impact of the Fourth Industrial Revolution. Some foresee large-scale unemployment and argue that this time will be different from the past. But there are also experts that say that this Fourth Industrial Revolution will have similar effects as past technological revolutions. Productivity will grow and newly created jobs will prevent rising unemployment.

However, the speed at which technological advances are happening is unprecedented. Moore’s law, which was coined by an Intel Co-Founder in 1965, states that the number of transistors in a dense integrated circuit doubles approximately every two years and thus predicts exponentially accelerating progress. It has been slowing in the last years, but there are still massive increases in computing power being made (McKinsey Global Institute, 2017, p. 27). Klaus Schwab, Founder of the World Economic Forum, argues that the changes are historic in terms of speed, impact, and scope. The exponential rather than linear development leads to an unprecedented velocity of breakthroughs. Both the development and diffusion of innovations is faster than ever. For example, some of today’s most successful companies such as Uber, Airbnb or Alibaba have only become ubiquitous in the last few years. The iPhone was only launched ten years ago, and already there were more than 2 billion smartphones globally in 2016 (Manyika & Chui, 2014). The fact that the Fourth Industrial Revolution builds on the Digital Revolution, combining a variety of technologies that disrupt society, the economy, and business, leads to an impact on all sectors. The entire system will be transformed, changing industries, business models, and our societal structures (Schwab, 2016, p. 2). One often cited example for wide-scale disruption is the comparison between Detroit in 1990 and Silicon Valley in 2014. In Detroit the three biggest companies had a market cap of USD 36 billion with 1.2 million employees. In Silicon Valley on the other hand, the market cap of the three biggest companies is USD 1.09 trillion while generating about the same revenues but while employing 10 times less people with only 137’000 employees (Manyika & Chui, 2014).

Like Schwab, in his TED Talk Erik Brynjolfsson says that “The Second Machine Age”, as he calls the Fourth Industrial Revolution, is characterized by exponential developments in the digital realm that are combinatorial (TED Talk). According to him, in the near future the effect of digital technologies will fully manifest through automation and the creation of “unprecedented things”. Computers are so versatile that it is impossible to predict what applications for them might be created in the next few years. Artificial Intelligence for example has made great leaps driven by the availability of huge amounts of data and the exponential increase of computing power. It is all around us e.g. in virtual assistants, translation software, or drones (Brynjolfsson & McAfee).

These developments lead to productivity being at an all-time high but also to a decoupling of productivity from employment, and of wealth from work, since a lot of activities can be automated. This will necessitate a rethinking of how society functions to deal with the arising challenges. Even voices saying that this time is similar to the other Industrial Revolutions, such as Ryan Avent in his book “The Wealth of Humans”, concur that we still have massive challenges to deal with, like in the First Industrial Revolution which brought about a whole new understanding of the state providing welfare, social security, and education and this revolution might be just as transformative and powerful, if not more so. The power today is to create mass prosperity of an unprecedented nature if we as a society take the right steps (Avent, 2016, p. 241).

In the past 200 years, we have seen massive improvements in almost all metrics that could be used to measure well-being. Since 1820, the share of people living in extreme poverty has decreased from 94% to 9.6%, and global child mortality fell from 43.3% to 4.25% (ourworldindata.org, 22.02.17), while global GDP (in current USD) more than doubled in the last 25 years alone (The World Bank, 22.02.17). However, we are starting to see some worrying trends. For example, in the US incomes have not kept up with this economic growth. A typical worker in the US earned USD 767 (in 2013 USD) per week in 1973 and that amount has been in decline since, now arriving at USD 664. Between 1949 and 1973, the US median household income grew in lockstep with per capita GDP, roughly doubling. If that had continued, the median would be at USD 90’000, however, it is only at 61’000, which is an increase of 22%. In the past, labor productivity moved in tandem with worker compensation, but in the last 30 years, we have been seeing an increasing divergence between these number, meaning that a bigger share of income goes to the holders of capital and business owners, and not the workers (Ford, pp. 35‒36). In selected G20 countries labor productivity increased by more than 18% while real wages only increased by 5% since 1999. Many economists such as John Maynard Keynes regarded it as a well-established rule that the fraction of national income going to capital and labor would remain constant over time. However, since the 1960s, the labor share declined from 63% to 58% in selected G20 countries and the trend is continuing. A declining labor share is often associated with rising income inequality since capital ownership tends to be concentrated in the hands of fewer individuals. There is data available by the OECD, which shows that the labor share decreased for the lower 99% while increasing by 20% for the top 1% in the last 20 years (International Labor Organization, 2015, pp. 3‒10).

These trends are bound to be further impacted by the automation of jobs and the invention of new technologies, thus allocating an even bigger amount to the owners of capital. This will pose a big challenge for us as a society in terms of how we distribute wealth and we might have to go through a turbulent phase like in past revolutions where we put new systems in place like the creation of the welfare state after the First Industrial Revolution. Before diving deeper into the impact on jobs, wages, and how to manage the changes, let us look at some of the technologies driving the change.

2.3 Some Context: Technological Drivers

According to Klaus Schwab, all new technologies have in common that they leverage innovations in the digital realm and in information technology. He groups the drivers of technological change in three main groups: physical, digital, and biological (Schwab, 2016, p. v). He argues that the merging and interaction of these clusters defines the Fourth Industrial Revolution. This section will give an overview over the three groups.

2.3.1 Physical

The physical realm includes some of the most visible innovations e.g. autonomous vehicles, the invention of new materials, and 3D printing. One particularly relevant development in manufacturing is the development of increasingly advanced robotics. Globally, there are currently approximately 1.5 million robots and this number could reach 25 million by 2025 according to McKinsey (Auschitzky et al., 2014). The cognitive abilities of these robots will only increase further and there are improvements being made in making robots more dexterous and improving their sensing abilities (Pluess, 2015, p. 10). Merrill Lynch projects that robots will perform 45% of manufacturing tasks by 2025, up from 10% in 2015 (Ma, Nahal, & Tran, 2015, p. 1).

2.3.2 Digital

The developments in the digital cluster build the basis for developments in other realms. The main area to highlight here are the advancements in Artificial Intelligence (AI), the area of computer science that concentrates on the development of systems that exhibit intelligence.  One sub-field worth mentioning is machine learning, which is a field in AI that develops systems which are able to learn and improve themselves. This provides opportunities for advanced analytics or big data. Companies have collected huge amounts of data in the past and AI allows them to use it effectively e.g. to automate complex cognitive tasks. Another big area of research are neural networks, which are AI systems that are based on simulating the way neurons interact in the brain (Manyika, et al., 2017, pp. 24‒25). Just this year, Elon Musk founded a new company, Neuralink, with which he is aiming to create a connection between our brains and computers, putting the time span until there is a meaningful interface at 5 years (The Economist, 2017). This might be optimistic, but it shows that research in these areas is moving fast and will likely lead to significant breakthroughs.

Another trend that has already been relevant for a while is the Internet of Things (IoT). It provides one of the main bridges between the physical and digital world, linking objects with virtual networks mainly by using sensors (Pluess, 2015, p. 12). Another development in recent years has been the Sharing Economy, with platforms connecting people and their assets, transforming the way we consume products and services.

One trend that could potentially disrupt the financial services is the Blockchain. This is a concept talked about a lot but understood by few. To simplify it, the Blockchain is a distributed ledger. When a transaction is made, a network of computers collectively verifies it, before it is recorded and approved. This technology is secure and shared, therefore no one single user can control or hack it and the contents can be inspected by everyone. It is currently used for recording Bitcoin financial transactions, but in the future it could be used to record marriage license, educational degrees, titles of ownership etc. (Schwab, 2016, pp 17‒19). In the second part of this thesis, which focuses on the financial services industry this technology will however not be the focus, since its wide-spread use still seems far away and the potential impact cannot yet be determined.

2.3.3 Biological

This category includes innovations in fields such as precision medicine and synthetic biology, the customization of organisms by writing DNA. This opens up a variety of ethical questions such as the debate around designer babies. However, we will not focus on these developments in this paper, since their impact on jobs is not as big as developments in the digital and physical clusters.

3 Impact of the Fourth Industrial Revolution on jobs

With the velocity, scope and depth of the development of the technologies discussed above, it is certain that the nature of work will change across industries and jobs. The main questions are to what extent automation will substitute for labor, how long it will take and how far it will go (Schwab, 2016, p. 35).

Up until now consumers have been the ones who benefited most from the Fourth Industrial Revolution. New services and products are being created that increase the efficiency and productivity of our lives basically for free. One can make a payment, listen to music, watch a movie, catch a taxi, or book an apartment and flight for a holiday remotely on one’s smartphone. A tablet has the processing power of 5000 computers 30 years ago, while the cost of storing information is near zero (Schwab, p. 11). However, in the future the supply side will also be increasingly disrupted by new technologies and become more productive and efficient. Labor will be substituted by machines in various activity, hopefully leading to better jobs to compensate the ones that are lost (Prisecaru, 2016, p. 61).  However, there have been indicators, that technology might not continue creating new and better jobs in the future. There is evidence that there are indeed better jobs being created, but also fewer ones than have been destroyed (West, 2015, p. 7).  A concern is that the labor market might become segregated into high-skilled jobs and jobs with low salaries, while the middle-skilled jobs are substituted. This might lead to higher inequality, which is a big concern since it has negative economic and social effects (Stiglitz, 2015). The owners of capital could be at an advantage to workers and the decline in labor share described above could accelerate.

This chapter will first examine two prevalent viewpoints on the impact of the Fourth Industrial Revolution: Some experts say that unemployment will soar and we will have to rethink the nature of work and others claim that the jobs that are lost will be replaced by new ones. The pessimists, who often have a background in the tech-industry say that this time is different and our jobs will be taken by machines, while the optimists, often historians and economists, think that technology will create more jobs than it destroys than in the past. A Pew Research Center study that asked almost 1’900 experts about the impact of emerging technologies, found that 48% of them envisioned a future “in which robots and digital agents have displaced significant numbers of both blue- and white-collar workers”. This means that experts’ opinions are almost split in half of where the development will lead (Smith & Anderson, 2014, p. 5)

But who will be right? This chapter will examine both arguments. For an overview of all studies cited in this chapter, please refer to Table XX in the Appendix.

3.1 Perspective A: Unemployment will increase dramatically

“If current trends continue, it could well be that a generation from now a quarter of middle-aged men will be out of work at any given moment.” This quote by former U.S. Treasury Secretary Lawrence Summers encapsulates the anxiety that some are feeling. He argues that providing enough work will be the most important challenge our society will face (West, 2015, p. 9). In some countries, there are already long-term trends towards a lower level of employment. In the US, the share of adult males in the labor force recently hit its lowest level since 1978 (The Economist, 2014).

As outlined above, in recent years there has been a revival of concerns that automation could lead to a jobless future (Arntz, et al., 2015, p. 4). Recent studies claiming that a substantial share of jobs are automatable have fueled these concerns. In a seminal 2013 study, Oxford researchers Frey and Osborne estimate that 47% of jobs in the US are susceptible to computerization at the current state of technology. Another report by Citi concludes that across the OECD, on average 57% are automatable and in China that percentage reaches an incredible 77% (Rahbari, et al., 2016, p. 11). A Deloitte report that replicated Frey and Osborne’s methodology puts the number for Switzerland at 48% (Brandes & Zobrist, 2015, p. 6).

Frey and Osborne took developments in machine learning, which impact cognitive tasks, and advanced robotics, which affect manual tasks, into account and categorized 702 occupations by probability of computerization. They differentiated between routine tasks, which follow explicit rules and can be executed by machines, and non-routine tasks, which cannot yet be expressed in computer code. In the past, job substitution has mainly been confined to manual and routine cognitive tasks. With advances in AI, big data, and other fields detailed in Chapter 2.3, automation is now also spreading to other tasks.  (Frey & Osborne, 2013, pp. 14‒16).

In the US, the most-common occupations are retail salesperson, food and beverage server, office clerk, and cashier. These occupations employ more than 15 million people or 10% of the labor force. According to Frey and Osborne each of these jobs is highly susceptible to automation (Thompson, 2015).

Companies such as Google, HP, Intel or Procter & Gamble have been collecting a staggering amount of data on their employees. E-mails, phone record, web searches etc. are often captured. Until now, this data had no real purpose except to assess employees or manage them more effectively (Peck, 2013). With the developments in big data and deep learning technologies, this data could be used to develop software that could automate a big share of the work performed. A question to ask oneself to determine whether one’s job could be automated in the near future, is whether someone could become proficient at it by repeating already completed tasks over and over. If that is the case, an algorithm might be able to do some if not all of that job (Ford, p. 93). Additionally, as top managers become more data-driven, the need for extensive teams of knowledge workers who collect and analyze information will decrease. This would lead to flatter organizations and many of the jobs of clerical workers, middle managers, and skilled analysts disappearing (Ford, p. 95).

Of course this sounds easier than it is. Machines still have a long way to go before they are able to perform complex task without supervision. Additionally, it might be difficult for companies to implement automation technologies because it might be more expensive than just relying on cheap labor. The IT infrastructure underlying processes might be old and not easily understood, and a lot of companies in sectors such as banking deal with a high level of complexity that few understand. The transformation of how they do business often requires a top-down approach and a real commitment to change, as Stephan Erni, a Manager at Bain & Company who specializes on the financial services, said in our interview (see Appendix for full transcript). Often consulting firms are hired to make sense of the complexity and build up a new IT infrastructure that leads to higher efficiency. Such fundamental changes are often prompted by crises or recessions that bring with them the pressure to reduce costs and the need for jobs restructurings.

The technological advances described above mean that jobs currently done by highly trained white-collar workers could be automated and not just manual work like we are used to. In the future, both white- and blue collar jobs will be equally vulnerable to automation. The key is still how routine a task is – according to Andrew Ng, a professor at Stanford University, a highly trained radiologist might have a higher probability of automation than his personal assistant, because he or she does many different things that cannot all be automated in the near future. The workforce is mainly split into two groups that hold non-routine jobs: the highly paid, high-skill workers such as managers on the one hand and he low-paid, unskilled workers such as servers in the fast-food industry on the other hand (The Economist, 2016). This has lead economists to increasingly worry about job polarization, with middle-skill jobs declining, while low-skill and high-skill jobs are expanding (Feng & Graetz, 2015, p. 32). The increase in employment in high wage occupations, where non-routine cognitive tasks are performed, can in turn raise the demand for non-routine manual tasks that are usually performed by low-wage occupations e.g. personal services. Studies showed that the fall of IT prices in 10 European countries led to the decline in the share of employment in middling occupations, while there was no adverse effect on the share of employment in the lowest paid occupations (Jerbashian, 2016, p. 3). Another study conducted in 16 Western European countries also showed that the occupational employment structure is polarizing both within and between industries (Goos, Manning, & Salomons, 2014, p. 2). This trend could continue in the future since most jobs especially in the middle-skill sector are on some level routine and predictable when broken down. Few people engage in truly creative work that cannot be replicated by a computer (Ford, p. xvii). Most jobs can be broken down into routine tasks on some level and in the future, just parts of the jobs could be automated. This could open the door to a reduction in job satisfaction from performing increasingly small parts of a task, similar to how the satisfaction of creating things was reduced by the use of interchangeable parts and deskilling in the 19^th century (The Economist, 2014). Of course, the opposite could be true, with “human” skills such as creativity and social skills becoming more important.

Either way, it has never been more important to have special skills and the right education and to have the ability to use the technology to create value. Conversely, there has never been a worse time to be a worker with only “ordinary” skills that can be easily acquired (McAfee).

A popular counter-argument to a jobless future is that the jobs will be replaced by new jobs like it has always been in the past. However, the Fourth Industrial Revolution appears to be creating fewer jobs in new industries. A report estimates that only 0.5% of the US workforce is employed in industries that did not exist 15 years ago. This percentage is far lower than it has been in the past with 8% of new jobs being created in new industries in the 1980s and 4.5% in the 1990s. (Schwab, p. 37). 90% of workers today are employed in occupations that existed 100 years ago and only 5% of jobs generated between 1993 and 2013 came from sectors such as computing, software, and telecommunications. Today’s new industries are more labor-efficient seem to raise productivity by making existing workers redundant and not by creating new products that need more labor to produce. For that reason, the historian Robert Skidelsky concluded that “sooner or later, we will run out of jobs”, comparing the exponential growth in computing with decidedly not exponential growth in the complexity of jobs (Thompson, 2015).

We are still far away from unemployment rates of 25% that Larry Summers predicts, and it is questionable whether it will ever get that far. But technology could continue exerting pressure on the value and availability of work. This could eventually lead to a new normal, where the expectation that work is a central feature of adult life vanishes for a substantial part of society (Thompson, 2015).

The new sectors created through technology also tend to favor a long-tail distribution of income, which could lead to higher inequality. These winner-takes all business models are central to companies that dominate the internet sector e.g. Google and Amazon making it harder for SMEs to exist (Ford, p. 76). Trends such as increasing job polarization and a decreasing share of national income going to labor instead of capital are worrying. In 2013, a typical production worker earned approximately 13% less than in 1973 in real terms, while productivity rose by 107% and living expenses have significantly increased. Already, the top 5% of households are responsible for 40% of spending while job remain the main mechanism distributing purchasing power to the consumer (Ford, pp. xi‒xvii). Thomas Piketty articulates the fear that we might be heading into a hyper-unequal society in his book “Capital in the Twenty-First Century”. According to him there could be a top 1% of capital-owners and “supermanagers” securing an increasing share of national income and accumulating even more capital. The rise of the middle class in the 20^th century was a very important political and social development and its decline could lead to more unstable and antagonistic politics (The Economist, 2014). We can see trends towards populism and a polarization of politics with the election of politicians such as Donald Trump in the US. This is not to say that growing inequality was the only reason for his election, but economic anxiety among white males was undoubtedly a factor.

There is the potential for real standards of living increasing even as wages stagnate if the prices of critical goods fall accordingly. However, this necessitates an evolution in our social safety net. When the wages of people without exceptional skills fall below a reservation wage, the wage at which people decide they are better off not looking for work, a larger share of people will decide not to work, which will pose an even bigger hurdle for our social security systems (Avent, 2016, p. 76).

But not only inequality within a country could be an issue but also between countries globally. Let us take China as an example, where 77% of jobs are automatable according to the Citi study cited above (Rahbari, et al., 2016, p. 11). In the past, many jobs have been off-shored from developed countries to countries such as China and almost 30% of the Chinese labor force are employed in industry, which mainly consists of manufacturing. The iPhone manufacturer Foxconn, one of the world’s largest 10 employers with approximately 1.3 million employees, has recently automated away 60’000 employees in just one factory and plans to have a 30% automation rate in its Chinese factories by 2020. The long-term plan is to fully automate their factories with only a handful of employees (Morris, 2016). A big trend is also there-shoring of manufacturing jobs when labor costs are no longer the key driver of competitiveness due to automation. Between 1990 and 2012, 1.2 million jobs in the US textile industry were offshored to low-wage countries, decimating the industry. In recent years, this trend has reversed. US textile exports rose by 37% between 2009 and 2012, being driven by automation and increasing wages in countries such as China (Ford, p. 9). Even though demographic trends and the rise of the services sector might mitigate this automation of manufacturing, it still poses a big challenge for countries such as China. In the US, where manufacturing currently makes up less than 10% of employment, factories approaching full automation only has a smaller effect than in countries such as China (Ford, p. 10). The Fourth Industrial Revolution could lead to the winner-takes-all principle might not just play out within countries but also globally between nations, which could lead to social tensions, conflicts, and a more volatile world (Schwab, p. 47).

There is definitely potential for dramatic change. It remains to be seen whether the pessimists will be right this time – there have always been labor-market doomsayers that have succumbed to the Luddite fallacy. However, there will be productivity gains. The question is whether they will concentrate in the hands of a few capital owners or be spread across the labor force e.g. by being used to create new businesses that in turn hire more labor. It is unclear whether this will aggregate enough to replace the jobs that are lost (The Economist, 2014). With all the disagreement on the extent of jobs loss there is a consensus in most studies that job polarization and inequality will increase (Acemoglu, 2000, p. 4). This could lead to the feedback loop between productivity, rising wages, and increased spending collapsing for good and levels of fragmentation, isolation, and exclusion across societies could rise (Gratton, 2011).

3.2 Perspective B: Substituted jobs will be replaced with new ones

After painting a rather pessimistic future, let us examine the counter-arguments optimists, who argue that just as in the past, new jobs will be created to replace the ones lost and that the occupations will just change. A strong advocate of this idea, the economist Robert Gordon argues that “recent progress in computing and automation is less transformative than electrification, cars, and wireless communication, and perhaps even indoor plumbing. Previous advances that enabled people to communicate and travel rapidly over long distances may end up being more significant to society’s advancement than anything to come in the twenty-first century” (West, 2015, p. 9). It is almost inevitable that there will be a profound impact on labor markets around the world, but this does not mean there has to be man-versus-machine dilemma. The technological advances we will experience might serve to enhance human labor and cognition and with the right education humans could work with and alongside machines (Schwab, p. 40). There are 9 main arguments that experts cite to alleviate the concerns discussed in Chapter 3.1 that we will look at here:

3.2.1 The occupation-based approach used by Frey and Osborne is flawed

The study by Frey and Osborne has obviously incited a lot of concerns, since 47% of jobs being susceptible to automation seems to be an extremely high percentage. But this is only alarming at first glance and it does not mean that in 20 years half of our jobs will be gone. As outlined above, the study and others like it follow an occupation-based approach, meaning that they assume that whole occupations rather than individual tasks will be automated. This might lead to an overestimation of how automatable jobs are, since even occupations that are categorized as high-risk still often have a significant share of tasks that are not easy to automate such as face-to-face interactions. An OECD study supports this view, taking the heterogeneity of tasks within a job into account and following a task-based approach instead, yielding much more conservative estimates. The study found that on average, across 21 OECD countries, 9% of jobs can be automated, with the lowest percentage being 6% in Korea and the highest 12% in Austria (Arntz, et al., 2015, p. 4). This makes the threat from technology seem much less pronounced, even though 10% is of course not a negligible share either.

A study conducted by the McKinsey Global Institute, which focused on 46 countries that represent approximately 80% of the global workforce, found that the percentage of jobs that can be fully automated at the current technology level is only 5%, which is even less than the OECD estimate since they did not just consider technical feasibility but also regulatory and societal issues. However, they also found that in middle-skill categories this percentage can rise to between 15% and 20%, which is another indicator for imminent job polarization. McKinsey pursued a similar approach as the OECD by examining more than 2’000 individual activities and quantifying their automation potential. Nonetheless, this does not mean that the impact of automation is negligible. Another finding of the study was that most jobs will be impacted by automation to some degree. According to the McKinsey Global Institute, approximately 60% of all jobs are made up of at least 30% of activities that are susceptible to automation by 2055 with the adoption of currently existing technologies. Thus, most jobs will change and employees will have to interact with technology to a greater extent. Globally, they estimate that 50% of the world economy or 1.2 billion jobs could be affected by the adaptation of current technologies (Manyika, Chui, Madgavkar, & Lund, 2016, p. 3). Especially tasks that include the processing and collection of data or physical activity in a predictable environment have a high potential for automation (Manyika, et al., 2017, p. 5). McKinsey’s long-term outlook is positive despite a lot of jobs being affected because they anticipate that there will be a rise in productivity brought about by automation, assuming that the people displaced by automation will find new employment. Their report says that automation could lead to a shift of the same magnitude as the long-term shift away from agriculture and decreases of manufacturing in the US (Manyika, et al., 2017, p. vii).

3.2.2 The potential for automation does not equal actual job substitution

Another argument is that the actual employment loss will not equal the percentage of jobs susceptible to automation. First of all, the adoption of new technologies is a slow process since there are economic, legal, and societal hurdles to overcome. Factors such as the cost of the development and deployment of technological solutions, labor market dynamics e.g. wages and availability of workers, the economic upside of the implementation of new technologies, and social and regulatory acceptance, are all relevant factors that influence the velocity of adaptation. Thus, the substitution through technology could take longer than expected (Manyika, et al., 2017, p. vii). In many cases the fact that a job can be automated will not lead to its substitution because relative costs also matter e.g. when the benefits of human labor outweigh the benefits and costs of machines (Pluess, 2015, p. 12). Secondly, workers can adjust to new technologies by switching tasks and might not be made redundant. And finally, new technologies also create additional jobs since the new technologies need producing and supervision, and higher productivity could lead to more labor demand. Thus, the OECD and McKinsey argue that automation is unlikely to cause large-scale unemployment (Arntz, et al., 2015, pp. 22‒25).

3.2.3 New industries and jobs will be created

There is no doubt that new jobs will be created through automation. For example, the new labor-saving technologies need to be designed, produced, and supervised, therefore creating demand in new sectors. Millions of jobs are anticipated to be created by technology. The information and communications technology (ICT) sector is a key growth driver in OECD countries, accounting for 5.5% of total value added in 2013, and 15–­52% of investments 2008 to 2013 (OECD, 2015, p. 84).

Some new jobs might not adhere to the typical picture of a jobs that is prevalent today but be created by new business models such as the sharing economy and the circular economy (Pluess, 2015, p. 15). There might be a human cloud where today’s professional activities are dissected into precise assignments and distinct projects that are carried out by workers located all around the world who are not affiliated with a specific company (Schwab, p. 47). This gig economy is already on the rise with Uber, Etsy, and TaskRabbit[1] counting millions of people offering their services. There might be a future of hyper-specialization where most of the workers can find a niche they excel at and be matched with buyers all over the world.

Automation could also create space for people to work in more emotive occupations and become artists, yoga instructors, or therapists. Such emotional work could be as important in the future as hammering metal was in the past. These jobs might get less respect at first since cultural norms change slowly (The Economist, 2014).

However, even though new, good jobs will be created it will not be scalable mass employment like in the Second Industrial Revolutions with mass assembly line. This poses a dilemma. The new forms of work will likely fulfill a maximum of two of the following conditions: high productivity and wages, resistance to automation, and the potential to employ massive amounts of labor. Employment will be limited by the abundance of labor and increasing automation (Avent, 2016, p. 66)

3.2.4 Complementary Effect

Up until now we have mentioned the substitution effect of new technologies multiple times, which comes into play when jobs get replaced by automation. However, there will also be a complementary effect since technologies might improve the competitiveness of a company through increased productivity. Thus, with lower costs and prices there is more demand for the product, which in turn leads to a higher labor demand, which could compensate for the labor-saving effect that technologies can have (Arntz, et al., 2015, p. 23)

A recent study analyzing the US workforce between 1982 and 2012 found that employment grew significantly faster in occupations that made more use of computers since the technology made some parts of the job faster, enabling employees to do the other parts better. Thus, the computer-intensive jobs displaced less computer-intensive ones and reallocated rather than displaced jobs (Bessen, 2015, p. 2). Another study showed that the creation of jobs in the ICT sector results in significant spill-overs into other industries and results in up to five further jobs such as lawyer or waiter being created in the local economy (Moretti, 2010, pp. 1–7).

A widely cited example is that of the introduction of the Automated Teller Machines (ATMs), which many thought would replace bank tellers. However, the total number of employees increased instead. Even though the number of employees per branch decreased from 20 to 13 between 1988 and 2004, the number of bank branches actually rose by 43% due to decreasing costs of running a branch. Thus, the number of employees rose but their work mix changed away from routing tasks to things like sales and customer service (Bessen, 2015).

Bank tellers became less like check-out clerks and more like salespeople, solving problems for customers and selling the new products. This is a more cognitively demanding jobs and a variety of skills. Thus, automating one part of a task does not make the other part unnecessary but rather more important and increases their economic value. This is called the O-Ring effect with reference to the Challenger space shuttle disaster, which exploded just because its O-Ring, a small gasket, was faulty. This means, that in a chain all links have to hold for the overall mission to be successful and thus the improvement of one link in the chain increases the value of improving others. Automating the weakest link of a human process thus strengthens the remaining human stages. A central economic mechanism of automation is that it raises the value of the tasks that only humans can do (Autor, 2016).

A recent study by the Centre for European Economic Research calculated that between 1999 and 2010 in Switzerland there were 103’000 jobs lost due to the substitution effect but at the same time more than 234’000 jobs created through the complementary effect. However, the complementary effect might be decreasing with 8% of employees in the 1990s were employed in occupations created by new technologies, as opposed to 4% in the 2000s and below 1% since then (Deloitte, 2017, p. 5).

Finally, it is also often true that substitution effects of new technologies are overrated and the complementary effects underrated (Autor, 2015, p. 6). Thus, it is always important to consider all aspects and potential impacts of technologies.

3.2.5 Lump of Labor Fallacy

Some of the people that think that automation will cause mass unemployment are succumbing to the lump of labor fallacy. This is the notion that there is only a finite amount of work to do and therefore that if you automate some of it there is less work available (The Economist, 2016).

As automation frees our time and the scope of what is possible there will be new services and products invented that occupy our time and increase consumption. As Thorstein Veblen, asociologist and economist said: “Invention is the mother of necessity” (Veblen, 1916, p. 314). If an average worker in 2015 wanted to attain the average living standard of someone in 2015 he would have to work only 17 weeks a year. David Autor notes that material abundance has never eliminated perceived scarcity. He calls this the Never Get Enough principle (Autor, 2016). Additionally, the complementary effect described above comes into play, increasing demand in existing jobs.

3.2.6 The potential of new technologies is often overestimated

As described in the introduction, there is a long history of fears that our jobs will be taken over by automation going back more than a 100 years. The development of technologies often did not live up to the hype regarding particular technologies. In 1956, when AI research was invented, a lot of experts in the field believed that a machine would become as intelligent as a human in less than a generation. Nowadays, the claims of experts in the field are not much different. As history shows, the potential of new technologies tends to be overestimated by experts in the field. More specifically, the comparative advantage of machines over workers in tasks involving power of judgement, common sense, and flexibility is often overstated (Pfeiffer & Suphan, 2015). For example, in Germany the digitalization of the manufacturing sector is only making slow progress. Industrie 4.0, a project introduced by the German government to spur interconnectedness and ditialization in the industry is still relatively unknown. A 2015 survey among German firms by the Zentrum für Europäische Wirtschaftsforschung (ZEW) concluded that only 18% of all firms are familiar with the concept of Industrie 4.0 and only 4% of firms have started projects in the field (ZEW, 2015). Of course, Industrie 4.0 might be more widely known now than two years ago, because the term Fourth Industrial Revolution was popularized by the WEF in the past years. But the utilization of technologies still has not reached the technological possibilities yet (Arntz, et al., 2015, p. 23). The prediction is that until 2020 progress in the manufacturing sector will continue to advance slowly (Graumann, Bertschek, & Weber, 2015, p. 7).

Additionally, it is very difficult to teach machines how to function in an environment without clear rules and unambiguous descriptions. This is called Polanyi’s paradox and presents a major obstacle to the evolution of AI. So far, achievements in AI have largely relied on a trial and error principle by combining big databases with software and immense computing power. This allowed machines to beat humans in highly controlled environments such as a game of chess or the game of Jeopardy. Thus, it might take a long time for computers to match human intelligence(Deloitte, 2017, p. 7).

3.2.7 Demographic effects

Another argument that is often brought up is that the demographic changes most of the world is experiencing might counteract automation with baby boomers exiting the workforce soon and declining birthrates and an aging society mean that peak employment will be reached in most countries within 50 years. Global birth rates have dropped by more than half since the 1960s. Thus, there are two factors to consider. There will be a labor shortage and the skill mismatch will worsen due to automation and new job requirements in the interface between man and machine (Strack, 2014). The decline in the share of the working-age population will cause an economic growth gap to open with approximately half of the sources of economic growth from the past fifty years disappearing. Thus, automation could compensate for this trend at least partly, injecting between 0.8% and 1.4% into global GDP annually. This is, if we assume that the worker substituted by automation rejoin the workforce to be as productive as they were in 2014. Thus, by 2065, automation could add productivity growth in the realm of 1.1 billion to 2.3 billion FTEs in the largest economies i.e. G19 and Nigeria without having an adverse effect on labor. This would ensure continued prosperity in advanced nations and continuing growth in developing ones (Manyika, et al., 2017, p. 15).

3.3 Representation of opinions in the media

Public Perceptions/Medienanalyse

3 Auslöser: Event (wie WEF), neuer Report z.B. Deloitte, Buch wie Rise of the Robots und Feature (z.B. Economist)

A lot more coverage: WEF one of the main topics, St. Gallen Symposium topic etc.

BIG in Schweiz Volksabstimmung?

In a sense, it was surprising that 2016 was the year in which the social threat posed by rapid technological progress became a topic of serious and widespread discussion. It was, after all, the first year since 2007 when all of the world’s advanced economies managed to grow and in which unemployment rates fell across the rich world. Yet those bright spots could not distract from other worries. Pay rises remain elusive for many workers, despite economic growth. Each day seems to bring new evidence of the massive economic disruption to come: from self-driving vehicles deployed by Tesla and Uber to experimentation with cashier-less shops by Amazon. But the biggest warning signs that something in society has gone awry were the political shocks of Brexit and Donald Trump. These votes seemed to reflect not just a resentment at the elites who have captured most of the gains from recent economic growth, but also the dawning realization that the economic and social marginalization of whole classes of people might, thanks to technological change, become a permanent feature of the landscape.

3.4 Consolidation of opinions

So who is right: the pessimist or the optimists? It will probably be something in between. AI will not cause massive unemployment unless the technological development vastly exceeds our expectations. But what it will do is increase the velocity of the existing trend of automation, therefore massively changing the labor markets and reducing employment in some sectors, like past technological disruptions have done. It will likely lead to increased inequality both between countries and within societies if mismanaged. Unlike in the past, the benefits of new technologies are not widely distributed and the median wages have fallen behind productivity growth. There will also be a skill mismatch that needs to be addressed, with some of the new high-skilled positions not being possible to fill by the previously low-skilled workers. It is unlikely that a 55-year-old Walmart cashier will be able to do a complex job that requires technological skills. It will be vital for companies and governments to provide worker opportunities to retrain and acquire new skills and for these workers to adapt the changing circumstances. Many workers will have to retrain multiple times in their lifetime as jobs are replaced by machines. Additionally, and improved education system that places more emphasis on the skills required in the future will be vital. In Chapter XX we will look at some options that governments and companies could use to confront these challenges. But first let us consider, which skills will actually gain in importance in the future.